Research On Design Method Of Wideband Filter Based On Neural Network

Due to the rapid development of wireless communication technology,the research of microwave technology and equipment is advancing in the direction of high performance,miniaturization and low cost,while its research and development and production time are becoming more and more urgent.In order to adapt to the development speed of the current technology,it is particularly important to establish a set of high-efficiency and high-precision wideband filter design methods.This paper mainly takes a parallel coupled line wideband filter loaded with two-stage cascaded open-circuit branches as the research object.Based on the design and research of wideband filters at home and abroad in recent years,artificial neural networks are used to construct a correct electromagnetic substitution model,and then realize the size and structure design of the wideband filter with high precision and high efficiency.The main research work of this paper includes the following three aspects:First,through the transmission scattering matrix network analysis of the parallel coupled line wideband filter structure loaded with two cascaded open-circuit branches,it is demonstrated that the Cameron method can be used to construct the generalized Chebyshev system function as the goal of designing the wideband filter.characteristic response function.First,the function expression of the wideband filter is calculated by cascading the transmission scattering matrix,and the system function of the wideband filter is obtained by using the conversion relationship between the transmission scattering matrix and the scattering matrix;secondly,the number of transmission zeros obtained by the analysis of the system function and position to construct a generalized Chebyshev system function;finally,the characteristic response curves of the two system functions are compared.The simulation results show that the characteristic response curves of the two system functions are in good agreement.Second,a wideband filter design method based on transmission scattering matrix and artificial neural network modeling is proposed.The design method takes the characteristic response curve of the wideband filter as the input variable,and uses the characteristic impedance value as the output variable to carry out the reverse modeling of the neural network,which can accurately and efficiently realize the same wideband filter design function as the traditional theoretical analysis method.At the same time,by using the trained neural network model,the wideband filter design can be realized under the design indicators of different bandwidths,different ripple coefficients and different transmission zero positions within the data acquisition range.The simulation results show that the neural network model has good accuracy and universality.Third,a wideband filter design method based on ADS and deep neural network technology modeling is proposed.This design method takes the actual influencing factors such as dispersion and parasitic of some circuits into consideration in the design of the wideband filter,and directly establishes the relationship between the characteristic response and physical size parameters of the wideband filter,and alleviates the structural discontinuity of the wideband filter.performance degradation problem.At the same time,there is no need for in-depth theoretical derivation or solving of nonlinear equations,which effectively improves the efficiency and practicability of wideband filter design.The simulation results show that the wideband filter after optimization of physical size parameters still has good passband and stopband characteristics in the whole frequency range.